Natural Language Interaction with Human and Computer Tutors

Johanna Moore, University of Edinburgh

23 May, 15:30
55.410 (UPF, Edifici Tànger, Campus de la Comunicació-Poblenou)

Abstract:
One-on-one tutoring with a human tutor is widely considered to be the
most effective form of instruction.  Despite considerable research
aimed at understanding what makes human tutoring so effective,
computer-based tutors still lag behind human tutors.  One major
difference between human tutors and current computer tutors is that
humans participate in unconstrained natural language dialogue with
students.  They can interpret student answers to open-ended questions
and respond to them appropriately.  In this talk, I will describe a
series of studies carried out with the BEETLE II system, which teaches
concepts in the domain of basic electricity and electronics using
natural language.  We compare human-human computer-mediated tutoring
with two computer tutoring systems based on the same materials but
differing in the type of feedback they provide to the student's
natural language explanations of domain concepts.  Our results show
that there are significant differences in interaction style between
human-human and human-computer tutoring, as well as between the two
computer tutors, and that different dialogue characteristics predict
learning gain in different conditions.

Johanna's bio:
Johanna D. Moore, FRSE, FBCS CITP, is Prof of Artificial Intelligence
and Director of the Human Communication Research Centre at the
University of Edinburgh.  Her research brings together theory and
practice from computer science, linguistics and cognitive science to
inform the development of natural language technologies for use in a
wide variety of applications, including tutorial dialogue systems,
assistive technology, recommender systems, and information extraction
from multimodal archives.  She received her PhD in Computer Science
from the University of California at Los Angeles in 1989, and then
joined the University of Pittsburgh as an Assistant Professor of
Computer Science and Research Scientist in the interdisciplinary
Learning Research and Development Center.  In 1994, she received a
National Science Foundation, National Young Investigator Award for her
work in tutorial dialogue.  She became a Fellow of the Royal Society
of Edinburgh in 2002, and the British Computing Society in 2006.  She
has served as President of the Association for Computational
Linguistics and Chair of the Cognitive Science Society.  Prof. Moore
is the author of over 150 publications, and Associate Editor of the
journals of Cognitive Science and Speech Communication.  She has served
as Program Chair for ACL, Cognitive Science, and the World Conference
on AI and Education.